Glossary: Difference between revisions
Juho Kunsola (talk | contribs) (+ = Media forensics = + Media forensics deal with ascertaining genuinity of media.) |
Juho Kunsola (talk | contribs) (sourced definition of = Generative adversial network = from Wikipedia into a {{Q}}. GANs are frighteningly good at faking 2D pictures of (non-)existing people.) |
||
| Line 32: | Line 32: | ||
= Digital sound-alike = | = Digital sound-alike = | ||
When it cannot be determined by human testing, is some synthesized recording a simulation of some person's speech, or is it a recording made of that person's actual real voice, it is a '''[[digital sound-alikes|digital sound-alike]]'''. | When it cannot be determined by human testing, is some synthesized recording a simulation of some person's speech, or is it a recording made of that person's actual real voice, it is a '''[[digital sound-alikes|digital sound-alike]]'''. | ||
---- | |||
= Generative adversial network = | |||
{{Q|A '''generative adversarial network''' ('''GAN''') is a class of [[w:machine learnin|g]] systems. Two [[w:neural network|neural network]]s contest with each other in a [[w:zero-sum game|zero-sum game]] framework. This technique can generate photographs that look at least superficially authentic to human observers,<ref name="GANnips" /><ref name="GANs">{{cite arXiv |eprint=1406.2661|title=Generative Adversarial Networks|first1=Ian |last1=Goodfellow |first2=Jean |last2=Pouget-Abadie |first3=Mehdi |last3=Mirza |first4=Bing |last4=Xu |first5=David |last5=Warde-Farley |first6=Sherjil |last6=Ozair |first7=Aaron |last7=Courville |first8=Yoshua |last8=Bengio |class=cs.LG |year=2014 }}</ref> having many realistic characteristics. It is a form of [[w:unsupervised learning|unsupervised learning]]]].<ref name="ITT_GANs">{{cite arXiv |eprint=1606.03498|title=Improved Techniques for Training GANs|last1=Salimans |first1=Tim |last2=Goodfellow |first2=Ian |last3=Zaremba |first3=Wojciech |last4=Cheung |first4=Vicki |last5=Radford |first5=Alec |last6=Chen |first6=Xi |class=cs.LG |year=2016 }}</ref>|Wikipedia|[[w:generative adversarial network|generative adversarial network]]}} | |||
---- | ---- | ||
Revision as of 00:51, 6 April 2019
Bidirectional reflectance distribution function
“The bidirectional reflectance distribution function (BRDF) is a function of four real variables that defines how light is reflected at an opaque surface. It is employed in the optics of real-world light, in computer graphics algorithms, and in computer vision algorithms.”
A BRDF model is a 7 dimensional model containing geometry, textures and reflectance of the subject.
The seven dimensions of the BRDF model are as follows:
- 3 cartesian X,Y,Z
- 2 for the entry angle
- 2 for the exit angle of the light.
Covert modeling
Covert modeling refers to both covertly modeling aspects of a subject i.e. without express consent.
Main known cases are
- Covertly modeling the human appearance into 7-dimensional [[#Bidirectional reflectance distribution function|]] model or other type of model.
- Covertly modeling the human voice
There is work ongoing to model e.g. human's style of writing, but this is probably not as drastic a threat as the covert modeling of appearance and of voice.
Deepfake
“Deepfake (a portmanteau of "deep learning" and "fake") is a technique for human image synthesis based on artificial intelligence. It is used to combine and superimpose existing images and videos onto source images or videos using a machine learning technique called a "generative adversarial network" (GAN).”
Digital look-alike
When the camera does not exist, but the subject being imaged with a simulation of a (movie) camera deceives the watcher to believe it is some living or dead person it is a digital look-alike.
Digital sound-alike
When it cannot be determined by human testing, is some synthesized recording a simulation of some person's speech, or is it a recording made of that person's actual real voice, it is a digital sound-alike.
Generative adversial network
“A generative adversarial network (GAN) is a class of g systems. Two neural networks contest with each other in a zero-sum game framework. This technique can generate photographs that look at least superficially authentic to human observers,[1][2] having many realistic characteristics. It is a form of unsupervised learning]].[3]”
Light stage
“A light stage or light cage is equipment used for shape, texture, reflectance and motion capture often with structured light and a multi-camera setup.”
Media forensics
Media forensics deal with ascertaining genuinity of media.
“Wikipedia does not have an article on w:Media forensics”
Synthetic terror porn
Synthetic terror porn is pornography synthesized with terrorist intent. Synthetic rape porn is probably by far the most prevalent form of this, but it must be noted that synthesizing consentual looking sex scenes can also be terroristic in intent and effect.
- ↑ Cite error: Invalid
<ref>tag; no text was provided for refs namedGANnips - ↑ Goodfellow, Ian; Pouget-Abadie, Jean; Mirza, Mehdi; Xu, Bing; Warde-Farley, David; Ozair, Sherjil; Courville, Aaron; Bengio, Yoshua (2014). "Generative Adversarial Networks". arXiv:1406.2661 [cs.LG].
- ↑ Salimans, Tim; Goodfellow, Ian; Zaremba, Wojciech; Cheung, Vicki; Radford, Alec; Chen, Xi (2016). "Improved Techniques for Training GANs". arXiv:1606.03498 [cs.LG].